텍사스 전력 신뢰성 위원회(ERCOT)는 AI 데이터센터에서 급증하는 메가와트 규모 부하에 대비해 전력망 운영을 스트레스 테스트하고 있다.

Texas has no shortage of proposed data center megawatts, but determining how many will actually materialize remains a far more difficult challenge. This distinction is now central to the Electric Reliability Council of Texas (ERCOT) as the state navigates an unprecedented wave of AI, hyperscale, and other large-load interconnection requests.

In June, ERCOT reported tracking more than 438 GW of proposed large loads, nearly 89% tied to data centers. By August 3, Governor Greg Abbott noted that ERCOT was evaluating approximately 474 GW of connection requests—roughly 90% from data centers and more than five times the grid’s record peak demand.

Neither figure forecasts what will ultimately be built, and that is precisely the point. ERCOT’s new Batch Zero process is imposing stricter boundaries on Texas’ massive development pipeline, requiring projects to demonstrate sufficient maturity, technical documentation, and financial commitment to secure a place in the transmission plan.

Simultaneously, emerging requirements for voltage ride-through and dynamic modeling are driving home a critical realization for the AI infrastructure industry: at the scale of hundreds of megawatts, a data center is no longer merely a passive customer at the grid’s edge. Its operational behavior directly impacts grid stability.

For developers, utilities, and investors, Texas is serving as a large-scale stress test to distinguish between announced AI campuses and executable infrastructure.

The sheer volume of ERCOT’s large-load queue often obscures how preliminary many of these projects remain. ERCOT’s April 2026 monthly report provided a clear snapshot: large-load applications totaled 445.8 GW through 2033, yet 321 GW had not submitted any studies to ERCOT. An additional 93.7 GW remained under review, while only 22 GW had satisfied the applicable Section 9.5 requirements.

Against that vast development funnel, ERCOT reported only 5.9 GW of observed energized large loads, with another 3.2 GW approved for energization but still non-operational. These distinctions carry significant weight.

A project may possess land, an announced capacity target, and an interconnection request, yet lack the engineering maturity, financing, transmission solutions, or customer commitments required to reach operation. When hundreds of gigawatts of proposed load compete for the same generation and transmission resources, treating every request as equally probable becomes counterproductive for both grid planning and data center development.

ERCOT has grappled with this forecasting challenge internally. Earlier this year, the agency considered applying a historical “realization rate” based on whether previously anticipated computational loads had actually energized. Instead, ERCOT recommended using Batch Zero-qualified loads as a more reliable foundation for transmission planning.

The underlying message is straightforward: requested megawatts are not equivalent to executable megawatts. This distinction will likely grow in importance well beyond Texas.

As Data Center Frontier reported earlier this month in “Texas Tightens Oversight of Data Center Development,” Batch Zero originated from Texas Senate Bill 6 and a broader state initiative to impose greater discipline on extraordinary large-load growth.

Approved by the Public Utility Commission of Texas (PUCT) in June, Batch Zero replaces repetitive, project-by-project analyses for qualifying loads of 75 MW and above with a coordinated, system-wide study.

This approach allows ERCOT to evaluate competing projects collectively, determine how much load specific grid segments can reliably support, and identify the transmission upgrades necessary to serve them.

ERCOT President and CEO Pablo Vegas described the initiative as a “fundamental shift in how ERCOT manages” large-load growth. ERCOT is the first U.S. independent system operator to implement such a batch process for major electricity consumers.

Though transitional, the process carries substantial industry implications. Data center developers can no longer rely solely on securing a promising parcel or initiating conversations with utilities. Technical studies, dynamic models, site control, financial commitment, and credible construction schedules are now essential criteria for advancing projects into a grid plan increasingly shaped by competing loads.

These shifts are already evident at the utility level. In its first-quarter earnings report, American Electric Power stated that AEP Texas holds 41 GW of Senate Bill 6-compliant contracted load additions through 2030, with data centers—including cryptocurrency mining operations—accounting for 88% of that total. At the time, AEP Texas’ active ERCOT pipeline comprised 78 data center requests totaling 52 GW.

AEP Texas’ stated qualification criteria require proof of financial capability or commitment, upfront construction funding, payment of an ERCOT study fee, verified site control, and full disclosure of intended generation sources and overlapping project requests.

Even 41 GW of contracted load does not guarantee that all capacity will energize according to customers’ current timelines. AEP explicitly notes that deployment schedules depend entirely on resource availability.

Nevertheless, it serves as a far more reliable development indicator than the raw queue totals. For the data center industry, this progression—requested, qualified, studied, contracted, allocated, and ultimately energized—is becoming an essential framework for evaluating claims of future capacity.

Batch Zero faced additional scrutiny on August 3, when Governor Abbott directed ERCOT and the PUCT to rigorously verify data center projects before permitting them to advance.

The governor instructed regulators to scrutinize project ownership, public incentives, projected power and water demands, onsite generation plans, cooling technologies, and community impact mitigation strategies. Projects that fail to meet state and regulatory standards risk denial of grid connection.

Consequently, ERCOT missed its original August 7 Batch Zero classification deadline. On August 20, the PUCT granted exceptions permitting ERCOT to proceed under a revised timeline. ERCOT stated it aims to issue conditional classifications to transmission and distribution utilities by August 31, triggering a dispute and reconciliation period before classifications are finalized.

Meanwhile, conditionally classified base loads may participate in quarterly stability assessments while awaiting final classification. This introduces another critical distinction in a market long accustomed to announcing massive development pipelines.

Simply entering the ERCOT queue no longer places a project on equal footing with one conditionally classified into Batch Zero. Furthermore, conditional classification does not guarantee transmission capacity, energization, or construction approval.

For developers competing for grid capacity, this added granularity will prove highly valuable, enabling clearer differentiation between robust, bankable projects and purely speculative proposals.

ERCOT’s oversight of large computational loads extends beyond determining connection capacity; it also addresses grid stability during sudden disconnections.

In June, ERCOT reported that four clusters of large loads lacking adequate voltage ride-through capability could trigger over 3.2 GW of load loss during a grid disturbance. ERCOT identified insufficient ride-through performance among large computational loads as the primary driver.

A sudden drop in demand of that magnitude can independently destabilize grid frequency. To mitigate this, ERCOT has developed stricter dynamic-modeling requirements under PGRR144, which specifically govern large loads and large computational facilities.

The pending proposal mandates voltage ride-through model-quality testing and requires PSCAD models of computational facilities to be validated against physical hardware tests. Any material facility modifications that could alter ride-through behavior will also trigger additional ERCOT review.

This marks a significant evolution in AI data center engineering. Uninterruptible power supply (UPS) systems, power conversion equipment, backup generation, control systems, and workload-protection schemes have historically been optimized primarily for facility continuity and IT equipment protection. At the scale now planned in Texas, however, their interaction with the transmission grid has emerged as a distinct engineering priority.

A 300-MW or 500-MW AI campus that disconnects instantaneously during a voltage disturbance behaves fundamentally differently from conventional load spread across thousands of residential and commercial customers. Put simply, the largest AI campuses are evolving into grid-scale electrical systems in their own right.

Batch Zero is intended as a transitional bridge. ERCOT’s long-term strategy involves establishing a recurring Batch 1+ and Comprehensive Transmission Planning framework that systematically integrates large-load interconnection with routine transmission planning cycles.

ERCOT has outlined the eventual system as one where projects must satisfy defined “seriousness” criteria before entering a batch. Subsequent studies would then allocate available capacity by year and pinpoint necessary transmission upgrades.

The current roadmap targets ERCOT Board consideration of the Batch 1+ framework in February 2027.